Plugin/Theme Disclaimer
Is this project open source? Yes
Is this project completely free? Yes
Is this project vibe-coded beyond the author’s ability to comprehend how it works? No
Community Directory: Link
Semantic Auto-Linker — suggest missing links and explore your vault as a semantic space
A local-first plugin for finding missing wiki-links, reviewing them before anything is written, and seeing how they would change your vault graph.
Hey everyone — I recently shipped Semantic Auto-Linker, an Obsidian plugin for people with useful notes but not enough cross-linking.
When I first started using Obsidian, it was a great place to put my university notes, but I did not get the connected graphs I kept seeing in showcases. My vault had useful information, but very little cross-referencing.
About a year ago, I asked in the Obsidian Discord whether there was a way to scan a mostly disconnected vault and use AI/embeddings to suggest links between related notes. At the time, there did not seem to be a smooth automated workflow for it, so I eventually built the plugin I was looking for.
What it does
Semantic Auto-Linker can:
- scan your entire vault, or the current note, and suggest missing links to other files
- let you review every suggestion before editing anything
- preview graph impact before and after applying links
- suggest links using titles, aliases, acronyms, and local semantic retrieval
- explore your notes and concepts in a semantic space, projected into 2D or 3D with PCA
- insert accepted links inline or into a footer section
- exclude noisy targets, ignore generic terms, or manually retarget suggestions
Semantic space explorer
One of the main features is the embedding explorer. It projects notes and extracted concepts into a semantic space, where related ideas appear close together even when they are not linked yet.
Obsidian’s graph view shows the links you already made. Semantic Auto-Linker also helps show the relationships your vault implies.
Local-first semantic matching
Semantic mode runs locally by default with a built-in embedding model. Ollama is also supported if you prefer to use your own local embedding model.
No cloud account is required. Suggestions are review-before-write, so the plugin does not modify your notes until you approve changes.
Performance
On my vault of around 500 Markdown files, it runs fairly quickly. Performance will depend on hardware, vault size, and whether semantic embeddings are already cached.
The linking quality still has room to improve, but I hope it is already useful for people with under-linked vaults like mine. I would love feedback, especially from people with larger vaults or workflows where manual linking has become tedious.
If this seems useful, a GitHub star would help me understand interest and make the project easier for other Obsidian users to discover.
